Bird sound spectrogram decomposition through Non-Negative Matrix Factorization for the acoustic classification of bird species

Diversity Monitoring of Coexisting Birds in Urban Forests by Integrating Spectrograms and Object-Based Image Analysis

In the context of rapid urbanization, urban foresters are actively seeking management monitoring programs that address the challenges of urban biodiversity loss. Passive acoustic monitoring (PAM) has attracted attention because it allows for the collection of data passively, objectively, and continuously across large areas and for extended periods. However, it continues to be a difficult subject due to the massive amount of information that audio recordings contain. Most existing automated analysis methods have limitations in their application in urban areas, with unclear ecological relevance and efficacy. To better support urban forest biodiversity monitoring, we present a novel methodology for automatically extracting bird vocalizations from spectrograms of field audio recordings, integrating object-based classification. We applied this approach to acoustic data from an urban forest in Beijing and achieved an accuracy of 93.55% (±4.78%) in vocalization recognition while requiring less than ⅛ of the time needed for traditional inspection. The difference in efficiency would become more significant as the data size increases because object-based classification allows for batch processing of spectrograms. Using the extracted vocalizations, a series of acoustic and morphological features of bird-vocalization syllables (syllable feature metrics, SFMs) could be calculated to better quantify acoustic events and describe the soundscape. A significant correlation between the SFMs and biodiversity indices was found, with 57% of the variance in species richness, 41% in Shannon’s diversity index and 38% in Simpson’s diversity index being explained by SFMs. Therefore, our proposed method provides an effective complementary tool to existing automated methods for long-term urban forest biodiversity monitoring and conservation.

Bioacoustic classification of avian calls from raw sound waveforms with an open-source deep learning architecture

The use of autonomous recordings of animal sounds to detect species is a popular conservation tool, constantly improving in fidelity as audio hardware and software evolves. Current classification algorithms utilise sound features extracted from the recording rather than the sound itself, with varying degrees of success. Neural networks that learn directly from the raw sound waveforms have been implemented in human speech recognition but the requirements of detailed labelled data have limited their use in bioacoustics. Here we test SincNet, an efficient neural network architecture that learns from the raw waveform using sinc-based filters. Results using an off-the-shelf implementation of SincNet on a publicly available bird sound dataset (NIPS4Bplus) show that the neural network rapidly converged reaching accuracies of over 65% with limited data. Their performance is comparable with traditional methods after hyperparameter tuning but they are more efficient. Learning directly from the raw waveform allows the algorithm to select automatically those elements of the sound that are best suited for the task, bypassing the onerous task of selecting feature extraction techniques and reducing possible biases. We use publicly released code and datasets to encourage others to replicate our results and to apply SincNet to their own datasets; and we review possible enhancements in the hope that algorithms that learn from the raw waveform will become useful bioacoustic tools.